🦸 Community Superhero

Our community superhero this week is Joe Alioto, senior specialist solutions architect at AWS. Joe is a phenomenal storyteller and hardcore builder, and his passion is reflected in his work. I’ve been following along with his journey on AI agents, MCP servers, and the observability of both of them - extremely fascinating (and important) stuff! Thank you Joe, for everything you do for the community!

💯 Spotlight

When Werner Vogels talks, people listen. He’s a wise man who sees things other people can’t. His article from last week talks about how development gets better with age and how experience in all forms makes you speed up as much as it makes you slow down. The post talks a bit about how and why the generative AI hype skyrocketed like it did and how we need to continue focusing on building safe, repeatable patterns and look before we leap. It’s a sobering reminder to not give into the hype and treat it like the next advancement in tooling - not a “must have” in every application you build.

🔥 My Favorite Content

It’s clear that Marcos Henrique spends a lot of time thinking and building with AI. His article last week, Building an agentic medical analysis system that actually thinks, is full of forward thinking goodness and practical implementation tips. Marcos goes through the details of an autonomous AI agent system, (correctly) classifies agents as “microservices with brains”, and talks about what the future holds and what is going to be important for developers in the next year. I love this and think he’s right on the money.

You may have seen that Andres Moreno and I are in the middle of a live series on AI agent design principles. Last week, we got together to talk about building user prompts with the right amount of context. Too much and you’re overspending and falling into traps like instruction dilution and recency bias. Too little and you’ll get unpredictable results. In our stream, we go through finding the right balance and discuss when to rely on tools to enrich context. And we did it with a fun use case where we made the agent decide what to do with a failed package delivery that had raw steaks inside. BONUS - Andres and I are moving to our own YouTube channel next week, away from the Believe in Serverless channel. Be sure to subscribe for some funny and helpful shorts and live streams.

Khawaja Shams always posts great, deep-thinking technical blog posts. Last week he published an article on Valkey’s official blog about memory performance between Valkey 8.1 and 8.0 at hyperscale. The post goes through memory consumption and throughput as he pushed 50 million entries into a sorted set. The results are very surprising, but in a good way. As a bonus, Khawaja also did the same comparison between Valkey 8.1 and Redis 8.2 🌶️ Great stuff!

Pubudu Jayawardana continued his short series on detecting EventBridge target failures with a post on using enhanced monitoring. This article explains what enhanced monitoring is, how it works, and how you can incorporate it into your apps. Enhanced monitoring is a new feature as of a couple of months ago, and it helps peek inside the black box that is event delivery in AWS. I find this extremely useful and practical for anyone relying on EventBridge for message delivery.

We had a great session from Eric Johnson last week on Serverless Office Hours about building intelligent agents on AWS. Eric gives us solid advice, like what you should do to follow a minimal implementation philosophy and what best practices to follow. He uses Kiro to build and test everything live, which makes for a fun watch.

💡 Tip of the Week

I found Paul Markham’s musings comparing the rise of AI to the introduction of the steam locomotive really interesting. It seems like we’re in a day and age with some revolutionary changes, but it’s not the first time the world has been rocked. Great little comparison here.

🐣 New Releases

Reminder - AWS News is the best source for AWS-related service announcements. For all releases and summaries of what happened, head over there!

I saw Luc van Donkersgoed say he thought pre:Invent season was here. For the uninitiated, that means a swathe of super cool releases leading up to AWS re:Invent. You tell me what you think!

Claude Sonnet 4.5 was released last week and immediately made available in Amazon Bedrock. I’ve been using it this past week, and it is AMAZING!

ECS Managed Instances was announced last week that offers an interesting blend of managed infrastructure with EC2. I feel like this is the first step of many to get us to a generalized compute service in the future that is as tunable, customizable, and managed as you want 🤔

There are a bunch of other new releases, but none of them serverless or AI related - I encourage you to visit AWS news for more updates!

Last Words

One particular quote from Werner’s article stuck out to me about gen AI: “it’s also been challenging because it wasn’t released like other technologies. No one educated users before release.” This is spot on and as a result we see tons of “best practices” under misguided pretenses. I’m not sure anyone has all the answers right now, but my advice here is to take a big step back, but your production glasses on, and take a look at what you’re building again. It might not be the best release candidate because you followed a random hello world example on the internet.

That’s my take on the week, but what’s yours?

What did I miss? What made you nod along (or 🙄)? Hit reply if you’re reading the email. Prefer socials? Ping me on Twitter, LinkedIn, or email.

Happy coding!

Allen